Review node-level allocation budget for CUST-WP-0071-T03

Fresh read-only railiance01 capture (06:08Z) via the Custodian collector, 7d
Prometheus usage, per-workload proposals with owners, and a node budget. The
rule-based proposal reconciles at +140m only with serialized release/build/HPA
events (-340m all concurrent); no procurement. Vergabe stays 60m provisional
pending RAPPS-WP-0014-T04. T03 stays in progress: 54 changes lack receiving
tasks in 26 owner repos.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 8340@bnt-lap001
Assistant-Session: 2e192b3d-b3e8-4403-94d4-d55a2c8b476c
This commit is contained in:
tegwick 2026-10-21 08:31:37 +02:00
parent e3e996fc95
commit e72c5de92f
13 changed files with 26468 additions and 3 deletions

View file

@ -8,7 +8,7 @@ status: active
owner: codex
topic_slug: railiance
created: "2026-10-19"
updated: "2026-10-19"
updated: "2026-10-21"
flavor: implementation
related: [CUST-WP-0071, VERGABE-WP-0019, RAPPS-WP-0014, RCLUSTER-WP-0014]
depends_on: []
@ -44,6 +44,7 @@ refreshing stale utilization metrics or inventing financial attribution.
id: RESOURCE-WP-0008-T02
status: progress
priority: high
blocking_reason: "Node budget and per-workload comparison done 2026-10-21; only the Vergabe target replay result (RAPPS-WP-0014-T04) remains to be inserted into the 60m provisional slot."
state_hub_task_id: "1f538441-8b45-5d37-aeb3-345204b75117"
```
@ -63,8 +64,9 @@ whole node budget fits. Do not infer human workflow acceptance.
```task
id: RESOURCE-WP-0008-T03
status: todo
status: progress
priority: medium
blocking_reason: "Proposal and node budget published 2026-10-21, but none of the 54 owner changes has a receiving task in its owner repo (26 repos listed in the review), the coulomb probe owner is unconfirmed, and the Vergabe slot awaits RAPPS-WP-0014-T04."
state_hub_task_id: "3e6e6a59-71a5-5048-b7a5-ed61416a442c"
```
@ -75,3 +77,28 @@ finished while actionable leftovers exist only in prose. All application of
accepted sizing stays with existing runtime owners and their reviewed deployment
paths. This record does not create a parallel scheduler or change financial
authority.
## Node-level review — 2026-10-21
`docs/evidence/2026-10-21-node-allocation-review.md` and its directory carry a
fresh read-only capture (06:08Z; identity 92.205.62.239 verified), a 7d Prometheus
capture, per-workload proposals and the node budget. The capture reused the
Custodian collector with `--output-dir`. Requests are 3605m, while measured
demand is container p95 966m/peak 1740m plus host/k3s at 677–871m. Demand fits;
reservations do not. The kubelet has no kube/system-reserved, the Knative HPA
ceilings could claim 1520m, and zero-request workloads hide demand.
The rule-based proposal (p95×1.5 / peak×0.5 / floors) lowers steady requests to
2400m. It reconciles at +140m only when releases, builds and Knative scale-out
are serialized; it is −340m short when they all coincide. No procurement. The
Vergabe line stays at 60m provisional, pending the target replay
(RAPPS-WP-0014-T04 slot).
T02 holds the evidence-backed comparison. It stays in progress only for the
Vergabe replay result: insert it into `build_review.py` and regenerate.
T03 cannot finish because every non-Vergabe change still needs a receiving task in
its owner repo. The per-repo list is in `2026-10-21-node-budget.md`. Creating
those records is the owners' or the Custodian's decision; this record does not
create them.